KnowledgeConfig
autogen.beta.agent.KnowledgeConfig dataclass #
KnowledgeConfig(store, expose_tool=True, write_event_log=True, compact=None, compact_trigger=None, aggregate=None, aggregate_trigger=None, bootstrap=None)
Groups knowledge-related Agent parameters.
The store is registered into context.dependencies[KnowledgeStore] so policies (e.g. WorkingMemoryPolicy, EpisodicMemoryPolicy) can read from it without an extra parameter. Everything else is a side effect of attaching a store; each is opt-out via its flag below.
| ATTRIBUTE | DESCRIPTION |
|---|---|
store | The backing knowledge store. TYPE: |
expose_tool | If True (default), the agent gets an auto-injected TYPE: |
write_event_log | If True (default), the agent persists its stream history to TYPE: |
compact, | Optional compaction strategy and its firing rules. TYPE: |
aggregate, | Optional aggregation strategy and its firing rules. Strategy failures emit TYPE: |
bootstrap | Optional custom bootstrap. None falls back to TYPE: |